Efficient hybrid placement generation method and system of concrete three-dimensional meso-model

Abstract

The application discloses a kind of efficient mixed delivery generation method and system of concrete three-dimensional mesoscopic model, it is related to mechanics and composite material numerical simulation technical field, including: determining aggregate gradation and carrying out interval discretization processing;Establish polyhedral aggregate geometric template library;Based on in-situ generation strategy, construct large aggregate skeleton;Based on global random sequence adsorption strategy, delivery medium aggregate;And based on space grid index mechanism, accelerate the delivery of small aggregate.The application uses the above-mentioned efficient mixed delivery generation method and system of concrete three-dimensional mesoscopic model, for the delivery of huge number of small aggregate, innovatively introduces background space grid index algorithm, the global collision detection complexity of traditional random sequence adsorption algorithm in the number of aggregate is exponentially increased O ( N 2 ), optimization is only needed in constant level detection O ( C ) in local neighborhood, so as to break the computing power bottleneck of high volume fraction mesoscopic model construction, realize the efficient delivery of mass aggregate.

Description

Technical Field

[0001] This invention relates to the field of computational mechanics and numerical simulation technology of composite materials, and in particular to an efficient hybrid application method and system for generating three-dimensional microscopic models of concrete. Background Technology

[0002] Concrete-like materials are essentially multiphase heterogeneous systems composed of aggregates, mortar matrix, and interfacial transition zones. While macroscopic mechanical experiments can characterize the overall average properties of the material, they struggle to capture the complex internal processes of stress wave propagation and reflection, as well as the initiation and evolution of microcracks. With the development of computational mechanics, numerical simulation methods based on micromechanics have become an important tool for studying the dynamic behavior of heterogeneous materials. By establishing numerical models that include microscopic components such as aggregates and matrix, the fundamental causes of material nonlinearity and rate sensitivity can be revealed more deeply.

[0003] In mesoscopic numerical modeling, the realism of the aggregate model and the efficiency of the placement algorithm are two core issues. Early studies often used round or spherical particles to simplify the aggregate model. Although this was computationally efficient, it ignored the irregular geometry of real aggregates and the resulting aggregate interlocking effect, leading to biases in the prediction of material shear strength and crack propagation path. In recent years, three-dimensional random polyhedral aggregate models based on Voronoi polygons or CT scan reconstruction have gradually become mainstream, and can more realistically reflect the mesoscopic topology of concrete.

[0004] However, existing modeling techniques still face significant challenges in terms of computational efficiency and geometric accuracy for 3D mesoscopic models with high aggregate content. Currently, the random sequence adsorption algorithm is the most commonly used aggregate placement method in mesoscopic concrete modeling. Its logic involves randomly generating aggregate positions and determining whether interference occurs. Although the logic is simple, when dealing with complex gradations with high volume fractions or a large number of tiny particles, the remaining available space inside the specimen becomes fragmented as placement progresses, leading to an exponential increase in the rejection rate of new aggregate placements. Traditional global traversal search strategies require interference detection with each placed aggregate, resulting in a time complexity as high as O(N²), causing the computation time to surge to an unacceptable level and becoming a computational bottleneck limiting the construction of large-scale mesoscopic models.

[0005] Furthermore, in terms of geometric shape control, existing polyhedron generation methods, lacking fine-grained pose control, are prone to generating artificially induced orientation textures in the model. For example, directly translating and calling the aggregate template can result in a large number of aggregates having the same orientation, disrupting the isotropy of the microstructure. Traditional random rotation methods based on Euler angles are not uniformly distributed on the sphere, easily leading to clustering at extreme points and failing to guarantee strict randomness.

[0006] In summary, existing technologies for generating three-dimensional microstructure models of concrete have the following problems:

[0007] 1) Distortion of aggregate geometry or low modeling efficiency: Simple sphere models cannot reflect the edge effect and mechanical interlocking of aggregates, resulting in inaccurate prediction of mechanical properties; while modeling methods based on real polyhedra are extremely costly to construct high volume fraction models in a limited time due to the lack of efficient collision detection mechanisms when dealing with massive amounts of aggregates.

[0008] 2) Traditional placement algorithms suffer from computational bottlenecks: Existing random sequence adsorption algorithms mostly employ global search strategies for interference detection. When simulating gradation systems containing thousands or even tens of thousands of small-diameter aggregates, the number of interference detections increases exponentially with the amount of aggregates, resulting in extremely low algorithm efficiency or even failure to converge, making it difficult to meet the needs of refined modeling.

[0009] 3) Lack of strict isotropic control: When adding irregular polyhedral aggregates, the lack of attitude control methods based on strict mathematical distribution can easily lead to non-physical artificial textures inside the model, affecting the accurate simulation of the macroscopic mechanical isotropic properties of materials. Summary of the Invention

[0010] The purpose of this invention is to provide an efficient method and system for generating three-dimensional microscopic models of concrete through hybrid application, thereby solving the problems mentioned in the background art.

[0011] To achieve the above objectives, this invention provides an efficient method for generating a three-dimensional microstructure model of concrete through hybrid application, comprising the following steps:

[0012] S1. Perform aggregate gradation design and interval discretization: Divide the continuous gradation curve of concrete aggregate into three characteristic particle size intervals: large aggregate, medium aggregate and small aggregate.

[0013] S2. Construct a polyhedral aggregate geometry template library: Based on Voronoi Thiessen polygon technology and shrinkage-fitting algorithm, generate an aggregate geometry prototype library with real morphological features;

[0014] S3. Generate large-size aggregates as a mechanical skeleton: In the computational domain, an in-situ generation strategy is used to construct a large aggregate skeleton to form an interference-free aggregate skeleton.

[0015] S4. Randomly add medium aggregates: Call medium aggregates from the polyhedral aggregate geometry template library and add them;

[0016] S5. Accelerate the addition of small aggregates: Construct a background space grid index based on the background space grid index mechanism to accelerate the filling of small aggregates.

[0017] Preferably, S1 includes: using an interval discretization strategy to divide the continuous gradation curve of concrete aggregate into three characteristic particle size intervals: large aggregate, medium aggregate, and small aggregate, and calculating the target quantity of each grade of aggregate based on the principle of conservation of volume fraction in each interval.

[0018] Preferably, S2 includes: first generating a Voronoi polyhedron in three-dimensional space, then introducing a shrinkage factor to radially shrink the vertices of the polyhedron towards the centroid, and controlling the aggregate volume through a bisection iterative strategy until the preset particle size requirement is met.

[0019] Preferably, S3 includes: directly and randomly scattering seed points within the computational domain to generate Voronoi cells, and performing in-situ shrinkage on each cell to directly obtain non-overlapping large aggregates;

[0020] Large aggregates do not require collision detection during the production process.

[0021] Preferably, step S4 includes: extracting a medium aggregate prototype from the aggregate geometry library based on the existing large aggregate, applying a random rotation perturbation to it based on the Gaussian matrix algorithm of singular value decomposition, and using a random sequence adsorption algorithm for placement and global geometric interference detection.

[0022] Preferably, the Gaussian matrix algorithm based on singular value decomposition in S4, which applies random rotation perturbation, includes: generating a matrix whose elements follow a standard normal distribution, performing singular value decomposition on it to obtain an orthogonal matrix, and correcting the determinant to ensure that the generated rotation matrix is ​​uniformly distributed on the three-dimensional rotation group.

[0023] Preferably, the construction of the background spatial grid index based on the background spatial grid index mechanism in S5 includes: discretizing the computational domain into a regular spatial grid, and registering the existing large and medium aggregates into the corresponding grid cell list.

[0024] Preferably, in the background space grid indexing mechanism in S5, the set value of the background grid size is greater than the maximum feature size of the small aggregate.

[0025] Preferably, in step S5: a local neighborhood search mechanism is used during the small aggregate delivery process, which only retrieves the list of registered aggregates in the grid where the aggregate to be delivered is located and its surrounding 3×3×3 neighborhood for local geometric interference determination, in order to replace the global traversal search.

[0026] A system for an efficient hybrid application method for generating three-dimensional microstructure models of concrete includes:

[0027] Memory, used to store computer programs and aggregate geometry template library data;

[0028] The processor executes computer programs and performs the following: initializing the computational domain and completing aggregate gradation design and interval discretization calculations; invoking in-situ generation strategies to construct large aggregate skeletons; invoking Gaussian matrix algorithms based on singular value decomposition and random sequence adsorption algorithms to add medium aggregates; constructing a background spatial grid index and accelerating the addition of small aggregates based on a local neighborhood search mechanism; and outputting the final generated three-dimensional microstructure model data of concrete.

[0029] Therefore, the above-mentioned efficient hybrid application and generation method and system for three-dimensional microscopic models of concrete has the following beneficial effects:

[0030] (1) Extremely high modeling and computation efficiency: For the massive amount of small aggregates to be placed, this invention innovatively introduces a background space grid indexing algorithm, which optimizes the global collision detection complexity O(N2) of the traditional random sequence adsorption algorithm, which increases exponentially with the amount of aggregates, to a constant level detection O(C) that only needs to be performed in the local neighborhood. This breaks through the computational bottleneck of building high volume fraction microscale models and realizes the efficient placement of massive aggregates.

[0031] (2) Realistic micro-geometric reconstruction: Compared with the traditional simplified sphere model, the present invention uses Voronoi polyhedron technology to reconstruct aggregates, which can realistically reflect the irregular geometric shape, edge effect and mechanical interlocking effect of natural crushed stone aggregates, laying a geometric foundation for subsequent accurate simulation of the shear strength and crack propagation path of concrete materials.

[0032] (3) Excellent model statistical properties: The present invention adopts a random rotation matrix generation algorithm based on singular value decomposition, which ensures that the attitude of aggregate in space is mathematically strictly uniform and random, effectively eliminating the orientation texture caused by artificial induction in traditional modeling, and ensuring the isotropy of the microscopic model.

[0033] (4) High filling rate and accurate gradation: By adopting a layered mixing strategy of "in-situ generation of large aggregates, random sequence adsorption algorithm for medium aggregates, and grid-accelerated placement of small aggregates", this invention successfully solves the problems of large particles occupying little space and small particles requiring excessive computation at high volume fractions. Experiments show that the actual volume of aggregates in each layer of the generated model is within ±2% of the target gradation error, and the aggregates are evenly distributed in the direction of gravity without obvious segregation.

[0034] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0035] Figure 1 The flowchart shows a method and system for efficiently generating a three-dimensional microstructure model of concrete using a hybrid application.

[0036] Figure 2 This is a schematic diagram illustrating the principle of Voronoi polyhedral aggregate generation in an embodiment of the present invention, wherein (a) represents the random seed point distribution, (b) represents the generated Voronoi polyhedral set, and (c) represents a single polyhedral aggregate.

[0037] Figure 3 This is a schematic diagram of the "shrinkage-fitting" process in the process of establishing the aggregate geometric template library in an embodiment of the present invention, wherein (a) is the shrinkage process of a single aggregate, and (b) is the generated small and medium aggregate library sample;

[0038] Figure 4 This is a comparative schematic diagram of random attitude control based on singular value decomposition in an embodiment of the present invention, wherein (a) is the unprocessed ordered attitude, (b) is the processed random attitude, and (c) is the random direction vector distribution.

[0039] Figure 5 This is a schematic diagram illustrating the principle of local collision detection based on spatial grid index in an embodiment of the present invention;

[0040] Figure 6 This is a visualization of the three-dimensional microstructure model of concrete generated using this method. Detailed Implementation

[0041] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0042] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0043] Example

[0044] Please see Figures 1-6 This invention provides an efficient method for generating a three-dimensional microstructure model of concrete through hybrid placement, comprising the following steps:

[0045] S1. Aggregate Grading Design and Interval Discretization To recreate the heterogeneous characteristics of concrete in the numerical model, the aggregate gradation curve must first be determined. Considering that directly simulating a continuous gradation curve would lead to extremely difficult geometric modeling, this embodiment adopts an "interval discretization strategy," dividing the continuous gradation curve of concrete aggregates into three characteristic particle size intervals: large aggregate (characteristic particle size 15-20mm), medium aggregate (characteristic particle size 8-15mm), and small aggregate (characteristic particle size 4-8mm). Based on the principle of conservation of volume fraction in each interval, the target quantity of aggregates in each grade is calculated. This reduces the algorithmic complexity of subsequent calculations. For example, in this embodiment, the target number of large aggregates is approximately 240, medium aggregates are approximately 500, and small aggregates are used as fillers, approximately 2630.

[0046] S2. Constructing a polyhedral aggregate geometry template library: To address the cost issue of real-time generation of massive aggregates, a library of aggregate geometry prototypes with realistic morphological features is generated based on Voronoi polygon technology and a shrinkage-fitting algorithm. First, Voronoi polyhedra are generated in 3D space. Then, a shrinkage factor is introduced to radially shrink the polyhedron vertices towards the centroid. A bisection iterative strategy is used to control the aggregate volume until the preset particle size requirement is met, thereby establishing an aggregate geometry library containing rich morphological features.

[0047] This method employs an "offline generation + online invocation" strategy, utilizing the Voronoi polygon technique to generate basic polyhedra, such as... Figure 2 As shown. For any seed point in three-dimensional space, its Voronoi cell is defined as the spatial region whose distance to that point is less than its distance to any other seed point. To obtain aggregate particles with true spacing, a "shrinkage-fitting" algorithm is introduced. For any generated polyhedron, its geometric centroid is calculated. Introducing contraction factor , will the vertex Radial contraction towards the centroid:

[0048] ;

[0049] Iterative adjustment using the bisection method The value is calculated until the error between the aggregate volume and the target particle size is less than a preset threshold (such as 10%), thereby constructing a geometric template library of large, medium and small aggregates containing rich morphological features.

[0050] S3. Generating Large-Aggregate as the Mechanical Framework: A large aggregate framework is constructed within the computational domain using an in-situ generation strategy, forming an interference-free aggregate framework. To avoid large particles having difficulty finding empty spaces during placement, an "in-situ generation" method is adopted: particles are directly and randomly scattered within the cylindrical computational domain. Seek 1 seed point to generate a global Voronoi partition. Perform in-situ shrinkage on each cell to directly obtain the partition. Large, non-overlapping aggregates, such as Figure 3 As shown. This step utilizes the mathematical segmentation property of Voronoi cells, resulting in aggregates that are naturally interference-free, eliminating the need for time-consuming collision detection, and maximizing space utilization.

[0051] S4. Randomly Dispatch Medium Aggregates: Medium aggregates are retrieved from the polyhedral aggregate geometry template library and dispatched. Based on existing large aggregates, medium aggregate prototypes are extracted from the aggregate geometry library. A uniformly distributed random rotation matrix is ​​generated using a Gaussian matrix algorithm based on singular value decomposition. Random rotation perturbations are applied to the matrix to adjust the aggregate orientation. A random sequence adsorption algorithm is then used for dispatching and global geometric interference detection.

[0052] To eliminate artificially induced orientation texture, a uniform random rotation matrix generation algorithm based on singular value decomposition is used, with the rotation effect as follows: Figure 4 As shown. The specific steps are: generate a 3×3 matrix whose elements follow a standard normal distribution. Singular value decomposition yields orthogonal matrices. , correct To ensure the determinant is 1 (pure rotation), we obtain the rotation matrix. :

[0053] ;

[0054] Will The algorithm is applied to the vertices of the aggregates to make their orientation completely uniform and random in space. Then, a random sequence adsorption algorithm is used to place the medium aggregates and perform global interference detection with the existing large aggregates.

[0055] The Gaussian matrix algorithm based on singular value decomposition applies random rotation perturbation as follows: generating a matrix whose elements follow a standard normal distribution, performing singular value decomposition on it to obtain an orthogonal matrix, and correcting the determinant to ensure that the generated rotation matrix is ​​uniformly distributed on the three-dimensional rotation group.

[0056] S5. To accelerate the addition of small aggregates, a background spatial grid index is constructed based on a background spatial grid indexing mechanism to accelerate the filling of small aggregates. This addresses the bottleneck of global search efficiency when adding massive amounts of small particles. The background grid size is set to be slightly larger than the maximum feature size of the small aggregates. The computational domain is discretized into a regular spatial grid, and existing large and medium aggregates are registered in the corresponding grid cell list. When attempting to add new aggregates, a local neighborhood search mechanism is used. Only the grid containing the aggregate to be added and its surrounding 3×3×3 neighborhood are searched for local geometric interference determination, replacing the global traversal search.

[0057] For the largest number of small aggregates (approximately 2600), traditional global search algorithms are too inefficient. This method introduces a background space grid indexing algorithm, the principle of which is as follows: Figure 5 As shown. The specific steps are as follows:

[0058] S51, Grid Division: Set the background grid size Slightly larger than the maximum feature size of the small aggregate (5.5 mm in this embodiment), the computational domain is discretized into... A three-dimensional index matrix.

[0059] S52. Object Registration: Register all large and medium aggregates generated in steps S3 and S4 to the corresponding grid cell list according to their bounding box range.

[0060] S53, Local Detection: When attempting to add a new small aggregate Pnew, first calculate its grid coordinates ( , , Only the list of registered aggregates within the grid and its 3×3×3 neighborhood is retrieved. .

[0061] S54. Interference Detection: Only applicable to... The aggregate in the mixture is subjected to precise geometric interference determination.

[0062] Through this mechanism, the number of candidate objects in a single collision detection is increased from the global number. O ( N Reduced to constant level O ( C This greatly improves computational efficiency.

[0063] The concrete microstructure model generated using the above method is as follows: Figure 6 As shown in the figure. Statistical analysis shows that the error between the actual cumulative volume of aggregates in each grade and the theoretical gradation curve is controlled within ±2%, and the aggregates are evenly distributed in the height direction, verifying the effectiveness and accuracy of the present invention in modeling complex gradations with high volume fractions.

[0064] A system for an efficient hybrid application method for generating three-dimensional microstructure models of concrete includes:

[0065] The memory stores the computer program and aggregate geometry template library data. The processor executes the computer program and performs the following: initializing the computational domain and completing aggregate gradation design and interval discretization calculations; invoking the in-situ generation strategy to construct the large aggregate skeleton; invoking the Gaussian matrix algorithm of singular value decomposition and the random sequence adsorption algorithm to add medium aggregates; constructing the background space grid index and accelerating the addition of small aggregates based on the local neighborhood search mechanism; and outputting the final generated three-dimensional microstructure model data of concrete.

[0066] Therefore, this invention employs the aforementioned efficient hybrid placement generation method and system for a three-dimensional microstructure model of concrete. For the large quantity of small aggregates to be placed, it innovatively introduces a background space grid indexing algorithm, reducing the complexity of global collision detection in traditional random sequence adsorption algorithms, which increases exponentially with the amount of aggregates. O ( N 2 This is optimized to a constant-level detection that only requires processing within a local neighborhood. O ( C This breakthrough overcomes the computational bottleneck in constructing high volume fraction microscopic models, enabling efficient deployment of massive aggregates.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for efficiently generating a three-dimensional microscopic model of concrete through hybrid placement, characterized in that, Includes the following steps: S1. Perform aggregate gradation design and interval discretization: Divide the continuous gradation curve of concrete aggregate into three characteristic particle size intervals: large aggregate, medium aggregate and small aggregate. S2. Construct a polyhedral aggregate geometry template library: Based on Voronoi Thiessen polygon technology and shrinkage-fitting algorithm, generate an aggregate geometry prototype library with real morphological features; S3. Generate large-size aggregates as a mechanical skeleton: In the computational domain, an in-situ generation strategy is used to construct a large aggregate skeleton to form an interference-free aggregate skeleton. S4. Randomly add medium aggregates: Call medium aggregates from the polyhedral aggregate geometry template library and add them; S5. Accelerate the addition of small aggregates: Construct a background space grid index based on the background space grid indexing mechanism to accelerate the filling of small aggregates; The S2 includes: first, generating a Voronoi polyhedron in three-dimensional space; then, introducing a shrinkage factor to radially shrink the vertices of the polyhedron towards the centroid; and controlling the aggregate volume through a bisection iterative strategy until the preset particle size requirement is met. S4 includes: on the basis of existing large aggregates, extracting medium aggregate prototypes from the aggregate geometry library, applying random rotation perturbation to them based on the Gaussian matrix algorithm of singular value decomposition, and using a random sequence adsorption algorithm for placement and global geometric interference detection. The Gaussian matrix algorithm based on singular value decomposition in S4 applies random rotation perturbation as follows: generating a matrix whose elements follow a standard normal distribution, performing singular value decomposition on it to obtain an orthogonal matrix, and correcting the determinant to ensure that the generated rotation matrix is ​​uniformly distributed on the three-dimensional rotation group. In S5: During the small aggregate delivery process, a local neighborhood search mechanism is adopted, which only retrieves the list of registered aggregates in the grid where the aggregate to be delivered is located and its surrounding 3×3×3 neighborhood for local geometric interference determination, in order to replace the global traversal search.

2. The efficient hybrid application and generation method for a three-dimensional microscopic model of concrete according to claim 1, characterized in that, S1 includes: using an interval discretization strategy to divide the continuous gradation curve of concrete aggregate into three characteristic particle size intervals: large aggregate, medium aggregate and small aggregate, and calculating the target quantity of each grade of aggregate based on the principle of conservation of volume fraction in each interval.

3. The efficient hybrid application and generation method for a three-dimensional microscopic model of concrete according to claim 2, characterized in that, The S3 includes: directly and randomly scattering seed points within the computational domain to generate Voronoi cells, and performing in-situ shrinkage on each cell to directly obtain non-overlapping large aggregates; Large aggregates do not require collision detection during the production process.

4. The efficient hybrid application and generation method for a three-dimensional microscopic model of concrete according to claim 3, characterized in that, The background spatial grid index construction based on the background spatial grid index mechanism in S5 includes: discretizing the computational domain into a regular spatial grid, and registering the existing large and medium aggregates into the corresponding grid cell list.

5. The efficient hybrid application method for generating a three-dimensional microstructure model of concrete according to claim 4, characterized in that, In the background space grid indexing mechanism in S5, the set value of the background grid size is greater than the maximum feature size of the small aggregate.

6. A system applied to the efficient hybrid application and generation method for a three-dimensional microstructure model of concrete as described in any one of claims 1-5, characterized in that, include: Memory, used to store computer programs and aggregate geometry template library data; The processor executes computer programs and performs the following: initializing the computational domain and completing aggregate gradation design and interval discretization calculations; invoking in-situ generation strategies to construct large aggregate skeletons; invoking Gaussian matrix algorithms based on singular value decomposition and random sequence adsorption algorithms to add medium aggregates; constructing a background spatial grid index and accelerating the addition of small aggregates based on a local neighborhood search mechanism; and outputting the final generated three-dimensional microstructure model data of concrete.

Images